En

Fa

b

a

ARRAA

KSKFE

1

tmumloobmSbat[

eyatdt

0d

Chemical Engineering and Processing 49 (2010) 51–58

Contents lists available at ScienceDirect

Chemical Engineering and Processing:Process Intensification

journa l homepage: www.e lsev ier .com/ locate /cep

xperimental investigation and modeling of steam cracking of Fischer–Tropschaphtha for light olefins

eng Wanga,b, Yuanyuan Xua, Jie Rena, Yongwang Lia,∗

State Key Laboratory of Coal Conversion, Institute of Coal Chemistry, Chinese Academy of Sciences, Taiyuan 030001, ChinaGraduate University of Chinese Academy of Sciences, Beijing 100049, China

r t i c l e i n f o

rticle history:eceived 8 February 2009eceived in revised form 20 October 2009

a b s t r a c t

The characteristics of product distribution and the kinetic model for predicting the yields of the majorproducts from steam cracking of Fischer–Tropsch (F–T) naphtha have been investigated in a pilot plantunder various conditions. An analysis of the experimental data suggests that the naphtha produced via the

ccepted 8 November 2009vailable online 12 November 2009

eywords:team crackinginetic modeling

low-temperature slurry-phase F–T process is an excellent feedstock for the production of light olefins,especially ethylene. For steam cracking of two F–T naphthas studied, ethylene is the primary productvarying from 36.89 to 41.83 wt%, and the total yield of valuable light olefins (C2H4, C3H6 and 1,3-C4H6)is not less than 60.34 wt% under the conditions estimated. The experimental product distributions couldbe satisfactorily predicted by use of a detailed molecular reaction scheme which consists of a first-order

secon

–T naphthathylene

primary reaction and 37

. Introduction

The commercial exploration of F–T technology in energy indus-ry has brought interest in researching and developing a kinetic

odel to predict product distribution in pyrolysis of synthetic liq-id fractions for light olefins [1]. Though great efforts have beenade to study steam cracking of hydrocarbons, naphtha and gas oil,

ittle reported literature is involved in F–T naphtha as a feedstockf steam cracking for light olefins [2–5]. An experimental researchf steam cracking of F–T naphtha for ethylene has been reportedy Dancuar et al. at a pyrolysis pilot plant [6]. However, the kineticodel for predicting the product distribution from the pyrolysis of

asol Slurry-Phase Distillate (SPD) naphtha was not discussed. It haseen proven that kinetic model for cracking individual hydrocarbonnd their mixtures is a powerful tool to predict product distribu-ion and optimize operating conditions in the ethylene production7,8].

Modeling of naphtha pyrolysis has been attempted at three lev-ls of sophistication. The simplest model correlated the product

ields with some parameters such as the cracking severity index [9],nd the other extreme examples are mechanistic models based onhe free radical reaction kinetics [10,11]. Though great progress ineveloping mechanism models has been made in the past decades,he detailed information on radical reaction process of naphtha is

∗ Corresponding author. Tel.: +86 351 7117140; fax: +86 351 7037461.E-mail address: ywl@sxicc.ac.cn (Y. Li).

255-2701/$ – see front matter. Crown Copyright © 2009 Published by Elsevier B.V. All rioi:10.1016/j.cep.2009.11.005

dary reactions.Crown Copyright © 2009 Published by Elsevier B.V. All rights reserved.

still insufficient. Due to this difficulty, several kinetic constants infree radical models usually need to be adjusted to minimize the dis-crepancy between the experimental and predicted data [12–14]. Toavoid those difficulties, molecular kinetic models were developedto replace the radical reactions. These models are relatively sim-ple but approximate to the true nature of the reactions. This kindof models has been successfully used as a tool for predicting theyields of main products and optimizing the pyrolysis process ofpure hydrocarbons as well as petroleum fractions [6,11,13,15–21].Van Damme and Froment [20] reported gross molecular reac-tion models both for typical naphtha and for lumped componentssuch as n-paraffin, isoparaffin and naphthene. Based on individ-ual molecule reactions, they also developed a more detailed modelcontaining 49 molecular reactions. However, neither the detailsof the reaction scheme nor the numerical values of the variousparameters such as the pre-exponential factors, the activation ener-gies, and the initial selectivities were disclosed. Kumar and Kunzru[21] proposed a kinetic model consisting of 22 molecular reac-tions which could satisfactorily predict the product distributionfor steam cracking of naphtha, kerosene and gas oil. To date, themolecular reaction models are still applied to predicting the prod-uct distributions and optimizing the steam cracking processes ofconventional mixtures and heavy petroleum fractions. However,these models were only developed for the naphtha derived from

petroleum.

In this work, two paraffinic naphtha fractions obtained fromthe low-temperature slurry-phase F–T process are investigated ina pilot plant. The aim is to investigate the effects of the crack-ing severity on the yields of various products. In order to predict

ghts reserved.

5 ering and Processing 49 (2010) 51–58

tm

2

2

erflcsaaotbtcrl

2

tapp6teTatflt

2

fdwbaTwidowueildttartc

Table 1Characteristics of F–T naphthas.

Nap-1 Nap-2

Density (g cm−3) 0.7011 0.7206Sulfurs (ppm wt) <1 <1Aromatics Nd Nd

ASTM distillation (◦C)IBP 56.7 78.310 88.9 116.650 117.8 146.290 152.0 184.4

ethylene, ethane, propylene, 1,3-butadiene and 1-butene, whereasthe amounts of products including hydrogen, acetylene, methylacetylene, propane, isobutene, 2-butene, n-butane, isobutane wererelatively small. In this work, the C3

− to propylene mass ratio was

2 F. Wang et al. / Chemical Engine

he distribution of the major products, a special molecular kineticodel for the F–T naphtha has been developed.

. Experimental

.1. Experimental apparatus

The experimental apparatus used has been described in detaillsewhere [22], so only a brief introduction will be given here. Theeactor is in the form of a coil with a length of 20.0 m. The feedstockow rate can be varied from 1.0 to 3.0 kg/h by means of an electroni-ally controlled pump. The heat is supplied by eight electric heatingections, each operating independently, so that any given temper-ture profile in the reaction zone can be readily reproduced. Therere forty thermocouples along the reactor for the measurement ofutside tube temperature, including eight set points for tempera-ure control of reacting gas. The feedstock and water fed separatelyy metric pumps to the reactor are preheated and vaporized inhe first section, then pass through the reaction zone and undergoomplicated chemical reactions. The effluent mixture leaving theeactor passes through a quenching system where the gaseous andiquid products and water are separated.

.2. Experimental program

Experiments were performed with two hydrotreated F–T naph-ha fractions supplied by SYNFUELS CHINA Co. Ltd. Preheatersnd the furnace temperatures were set to desired values by usingroportional-integral-derivative (PID) digital controllers. The tem-eratures of water and naphtha in preheated section were set at00 ◦C. By the temperature control, different coil outlet tempera-ure (COT) values and temperature profiles could be achieved. Thisnabled different reaction conditions to be applied to the reactor.he flow rate of the feedstock varied from 2.00 kg/h to 2.50 kg/h,nd that of vaporized water from 1.00 kg/h to 1.30 kg/h. Residenceimes were obtained by varying the feed flow rate and the steamow rate. The coil inlet pressure (CIP) was kept constantly over allhe runs.

.3. Gas and liquid analysis

The detailed analyses required for gaseous products were per-ormed on a HP6890 gas chromatograph with a flame ionizationetector (FID) and a thermal conductivity detector (TCD). The TCDas used for the analysis of the amounts of hydrogen, methane, car-

on dioxide and carbon monoxide in gaseous samples. The FID waspplied to the analysis of organic compounds in gaseous products.he contents of C6-C8 monoaromatics in the pyrolysis productsere analyzed by another HP6890 gas chromatograph with a flame

onization detector (FID) and a DB-waxetr column, 0.32 mm iniameter and 60 m in length. The identification of the componentsf the saturated alkanes in the feedstock and the liquid productsas performed by an integrated GC-MS system (Agilent 5973),sing D-MS capillary chromatography column, 0.32 mm in diam-ter and 60 m in length. In addition, the liquid products collectedn each run were separated into three fractions: pyrolysis gaso-ine consisting of C5 fraction to distilling below 180 ◦C, pyrolysisiesel having a boiling range of 180–204 ◦C and pyrolysis oil dis-illing above 204 ◦C. The yields of gaseous and liquid products and

he concentration of the components in the product mixtures werenalyzed with a standard deviation of less than 5%. The total mate-ial balance of hydrocarbons determined in this way amountedo 95% or more. No measures were taken to analyze quantities ofarbon formed.

FBP 183.6 208.2

Average molecular weight 116.2 134.5

Nd: not detectable.

3. Results and discussion

3.1. Characterization of F–T naphtha

Two hydrotreated F–T naphtha fractions were investigated. Thefirst, referred to as Nap-1, had a boiling range of 56.7–183.6 ◦C(ASTM) and a density of 0.7011 g/cm3 (15 ◦C). The second, referredto as Nap-2, had a boiling range of 78.3–208.2 ◦C (ASTM) and adensity of 0.7206 g/cm3 (15 ◦C). Table 1 shows the most impor-tant characteristics of the F–T naphthas studied. Fig. 1 illustratesa complete analysis of Nap-1. As shown in Fig. 1, normal paraf-finic components, corresponding to main peaks, are in significantamounts of these which can be detected. The results of the qual-itative and quantitative analysis show that Nap-1 is a light F–Tnaphtha having more normal paraffins from C6H14 to C9H20,whereas Nap-2 is a heavy with a considerable amount of normalparaffins from C7H16 to C10H22. The main components are normalparaffins, 2-methylparaffins and 3-methylparaffins which accountfor more than 94.81 wt% of Nap-1 and 92.89 wt% of Nap-2. Fewaromatics and olefins are detected in both samples. The naphthafractions presented here can be described as a highly paraffinicproduct with few aromatic components.

3.2. Product distribution

The effect of the cracking severity on the product distributionwas investigated under the following range of operating condi-tion: COT, 790–850 ◦C; feedstock flow rate, 2.00–2.50 kg/h; waterflow rate, 1.00–1.30 kg/h; and CIP, 1.4 MPa. At the present exper-imental conditions, the major gaseous products were methane,

Fig. 1. Gas chromatography of Nap-1.

F. Wang et al. / Chemical Engineering and Processing 49 (2010) 51–58 53

ubsaapi

ttsiordimbcF

Fig. 4. 1-C4H8 yield vs. C3−/C3H6 mass ratio in F–T naphtha cracking.

Fig. 2. C2H4 yield vs. C3−/C3H6 mass ratio in F–T naphtha cracking.

sed as a severity factor to characterize the process conditionsecause it was the best measure of the severity of operation forteam cracking of complex feedstocks such as naphtha, kerosene,nd gas oil [23]. The C3

− fraction contains hydrogen, methane, C2nd C3 components. Characteristic results of the yields of variousroducts plotted against the C3

− to propylene mass ratio are shownn Figs. 2–7.

From the analysis of all the cases for the two F–T naphtha frac-ions, ethylene was the primary product varying from 36.89 wt%o 41.83 wt%. The favorable influence of increasing the crackingeverity on the conversion of two samples was reflected by anncreased concentration of ethylene in the final products. The trendf the ethylene yield plotted against the C3

− to propylene massatio is shown in Fig. 2. Methane and propylene were also pre-ominant components with a weight percentage of about 29.03%

n the cracked mixtures. With increasing the cracking severity, the

ethane amount increased monotonically (see in Fig. 3a). The 1,3-

utadiene yield passed through a maximum with an increase inracking severity, and the maximum yield was less than 5.40 wt%.or butene fraction, 1-butene was a major product. Due to the Fig. 5. H2 yield vs. C3

−/C3H6 mass ratio in F–T naphtha cracking.

Fig. 3. (a–d) Yields of gaseous hydrocarbons vs. C3−/C3H6 mass ratio in F–T naphtha cracking.

54 F. Wang et al. / Chemical Engineering and Processing 49 (2010) 51–58

+ vs. C

sbsataig0

Ctasiim

usitzpctoTibiawk

i

Fig. 6. (a and b) Yields of light olefins and C5

econdary pyrolysis and polymerization reactions, the amount of 1-utene decreased remarkably when the cracking condition becameevere (see in Fig. 4). In all runs, saturated light hydrocarbons werelso observed (mainly ethane, propane and butane). All the reac-ions forming these saturated hydrocarbons were inhibited for onlymonotonic decrease was observed when the cracking severity

ncreased (see in Fig. 3b–d). Among the gaseous products, hydro-en was present in small quantities in all runs, less than or equal to.79 wt% as shown in Fig. 5.

The total yield of valuable light olefins (C2H4, C3H6 and 1,3-4H6) varied from 60.34 wt% to 62.90 wt% (see Fig. 6a). Due tohe larger and lighter normal paraffinic content of Nap-1, the totalmount of light olefins was much higher in the final products. Fig. 6ahows that the light olefin yield passed through a maximum withncreasing the C3

− to propylene mass ratio. In other word, the max-mum light olefin amount can be obtained under the C3

−/propyleneass ratio of 4.60 for Nap-1 and 4.47 for Nap-2 in this work.Among the aromatic hydrocarbons identified in the liquid prod-

cts, benzene was a primary constituent. Under the same crackingeverities, Nap-2 yielded more benzene than Nap-1. Since aromat-cs were initially absent in the feedstock, they were formed duringhe steam cracking process. The amount of C6-C8 aromatics (ben-ene, toluene, xylene, ethylbenzene and styrene) in the pyrolysisroducts was less than 4.80 wt%, even at the most severe pyrolysisondition that the C3

− to propylene mass ratio was about 5.06 inhe case of Nap-2. Fig. 7 shows the influence of cracking severitiesn the yields of benzene and BTX (benzene, toluene and xylene).he yields of benzene and BTX increased with increasing the crack-ng severity, and the maximum amount was within 4.24 wt% fromoth samples. Among the aromatic products, the higher aromat-

cs (C9-C11) such as alkylated benzenes, indanes, and indenes were

lso found. No condensed aromatics with higher molecular massere observed, such as biphenyls and its homologues, which are

nown to be precursors to coke.Except for the C6-C8 aromatics, the liquid fraction was analyzed

n detail and separated into three fractions: pyrolysis gasoline,

Fig. 7. (a and b) Yields of C6H6 and BTX vs. C3−/

3−/C3H6 mass ratio in F–T naphtha cracking.

pyrolysis diesel and pyrolysis oil. Pyrolysis gasoline with the com-position of the C5 fraction to distilling below 180 ◦C was dominatingin the liquid product. It contained a significantly high content ofolefins, especially at low cracking severities. With an increase inthe cracking severity, the yields of higher olefins decreased, as wellas the formation of the gasoline fraction. The trend of the pyrolysisdiesel and the pyrolysis oil was contrary to that of the pyrolysisgasoline. The total amount of them was within 2.47 wt% under theexperiment conditions estimated. As expected, the steam crackingof Nap-2 yielded more heavy fractions due to the larger and heav-ier paraffinic content of it (see in Fig. 6b). Moreover, the C5

+ yieldfrom Nap-2, first decreased, and then increased with an increasein the cracking severity. However, the yields of secondary productssuch as aromatic hydrocarbons monotonically increased as shownin Fig. 7. The reverse trends of the decomposition of heaver olefinsand the formation of aromatics could partially explain the presenceof minimum in the plots of the C5

+ yield vs. the cracking severityfor Nap-2.

Figs. 2–7 well indicate that the F–T naphtha composition hasan influence on the yields of various products. The more amountsof light normal paraffinic hydrocarbons of the naphtha favors thehigher yields of the components of interest such as ethylene. Due tothe larger content of lighter paraffins in Nap-1, the C5

+ yield is lowerthan that from Nap-2 under the similar conditions. Comparing theyield of ethylene with that of the C5

+ fraction at various severities(Figs. 2 and 7), it could be expected that the ethylene yield from thetwo fractions would be much higher at optimum conditions. So thenaphtha used herein can be regarded as an excellent feedstock forproducing ethylene.

3.3. Kinetic model

The model of molecular reactions was initially used as publishedin literature [21,28]. However, a complete discrepancy between thepredicted and experimental data was observed in the simulationof F–T naphtha steam cracking. Therefore, a new kinetic model

C3H6 mass ratio in F–T naphtha cracking.

F. Wang et al. / Chemical Engineering and Processing 49 (2010) 51–58 55

Table 2Reaction scheme for steam cracking F–T naphtha.

No. Reaction equations Ea (kJ/mol) A (s−1)

1 Naphtha → 0.216H2 + 0.476CH4 + 1.631C2H4 + 0.221C2H6 + 0.665C3H6 + 0.168C3H8 + 0.005i-C4H10 + 0.002C4H10 + 0.280C4H8 + 0.072C4H6 + 0.003C4’sa

2 C2H6 → C2H4 + H2 247.26 3.557 × 1012

3 C2H4 + H2 → C2H6 238.30 (1.087 × 1013)b

4 C3H6 → C2H2 + CH4 261.86 9.854 × 1013

5 C2H2 + CH4 → C3H6 137.75 (1.802 × 107)b

6 C2H2 + C2H4 → C4H6 193.20 (2.700 × 107)b

7 2C2H6 → C3H8 + CH4 268.94 (2.032 × 1014)b

8 C2H4 + C2H6 → C3H6 + CH4 262.81 (9.773 × 1010)b

9 C3H8 → C3H6 + H2 214.99 3.394 × 1013

10 C3H6 + H2 → C3H8 153.18 (1.881 × 109)b

11 C3H8 → C2H4 + CH4 221.51 4.979 × 1011

12 C3H8 + C2H4 → C2H6 + C3H6 242.40 (1.649 × 1012)b

13 2C3H6 → 3C2H4 269.55 (2.912 × 1013)b

14 C2H4 → C2H2 + H2 209.55 6.419 × 109

15 C2H2 + H2 → C2H4 287.78 (2.090 × 1012)b

16 C3H6 → C3H4 + H2 204.29 2.359 × 109

17 C3H4 + H2 → C3H6 257.51 (1.822 × 1013)b

18 2C3H6 → 0.105CnHm + 0.487C5+ + 2.009CH4 235.18 (5.372 × 1011)b

19 C3H6 + C2H6 → 1-C4H8 + CH4 246.42 (1.767 × 1010)b

20 n-C4H10 → C3H6 + CH4 238.70 2.518 × 1013

21 n-C4H10 → 2C2H4 + H2 291.02 6.698 × 1012

22 n-C4H10 → C2H4 + C2H6 251.82 2.604 × 1010

23 n-C4H10 → 1-C4H8 + H2 280.36 5.814 × 1012

24 1-C4H8 + H2 → n-C4H10 154.71 (2.694 × 108)b

25 1-C4H8 → 2-C4H8 69.47 5.149 × 103

26 2-C4H8 → 1-C4H8 72.21 1.727 × 104

27 i-C4H10 → i-C4H8 + H2 213.22 2.927 × 108

28 i-C4H8 + H2 → i-C4H10 230.08 (4.663 × 1011)b

29 C3H6 + H2 → C2H4 + CH4 211.57 (2.249 × 108)b

30 i-C4H10 → C3H6 + CH4 229.70 5.118 × 1012

31 1-C4H8 → 0.105CnHm + 0.466C5+ 215.28 1.752 × 1011

32 1-C4H8 → H2 + C4H6 209.36 9.046 × 1010

33 H2 + C4H6 → 1-C4H8 113.73 (1.425 × 103)b

34 C2H4 + C4H6 → B + 2H2 144.29 (6.804 × 106)b

35 C4H6 + C3H6 → T + 2H2 144.32 (9.722 × 104)b

36 C4H6 + 1-C4H8 → EB + H2 243.84 (4.286 × 1011)b

37 C4H6 + C4H6 → ST + 2H2 125.09 (2.177 × 106)b

38 C4H6 + 2-C4H8 → X + 2H2 114.67 (1.858 × 106)b

oeffic

: styr

(cfipbair

cp

a

tatnittfi

a The stoichiometric coefficients for Nap-1; the stoichiometric c0.004, 0.314, 0.102 and 0.052, respectively.

b Units: m3/(mol ); B: benzene; T: toluene; EB: ethylbenzene; ST

Table 2) was developed by an analogy taken from modeling ofonventional naphtha using molecular reaction schemes. It wasrstly assumed that the F–T naphtha was represented as a pseudo-ure compound and its primary decomposition was representedy a first-order reaction. The selectivities of primary products weressumed to be constant. Then, the primary products were undergo-ng secondary reactions which were also represented by moleculareactions.

The first-order rate constant for the primary reaction was cal-ulated using a pseudoisothermal approach as discussed later. Therimary reaction can be represented as

Naphtha → a1H2 + a2CH4 + a3C2H4 + a4C2H6 + a5C3H6 + a6C3H8

+ a7i-C4H10 + a8C4H10 + a9C4H8 + a10C4H6 + a11C4’s

where a1, . . ., a11 are the stoichiometric coefficients, and C4’s isn olefin mixture consisting of C5

+ olefin products.In the steam cracking of F–T naphtha, enormous secondary reac-

ions can take place between the various primary products. Anttempt has been made to include the important secondary reac-ions that will account for the various products formed in F–T

aphtha pyrolysis. That is, the choice of the appropriate reactions

s based on the product distribution obtained from the experimen-al data with the guideline of radical mechanisms. In our study,he secondary reactions proposed by Kumar and Kunzru [21] wasrstly considered. This set of secondary reactions could satisfacto-

ients for Nap-2 are 0.318, 0.643, 1.900, 0.179, 0.721, 0.139, 0.005,

ene; X: xylene.

rily predict the product yields from steam cracking of petroleumnaphtha, kerosene and gas oil. However, with this scheme the pre-dicted ethylene and propylene values were lower and the predictedBTX value was much higher than our experimental values. Thus, theset of secondary reactions was not directly applicable for simulat-ing the F–T naphtha steam cracking. We modified this model byadding the reactions of methyl acetylene, isobutane and isobuteneto account for their formation in the final products. These probablereactions were based on the reaction model proposed by Sundaramand Froment [16]. For the explanation of the acetylene forma-tion, we used the dehydrogenation of ethylene to acetylene as arepresentative. The reversible reactions of 1-butene and 2-butenewere used as the representatives of isomerization reactions andaccounted for the 2-butene formation.

In the final set of secondary reactions (see in Table 2), the dehy-drogenation reactions of saturated hydrocarbons (C2-C4) representthe formation of light olefins (for example reactions 2, 9 and 23). Thehydrogenation reactions of unsaturated light hydrocarbons (reac-tions 3, 10, 24) are modeled on the formation of ethane, propaneand butane in the real reactions. Reactions 11, 20–22 representthe unimolecular decomposition of 1-propyl and butyl radicals to

yield ethylene, propylene and methane. Reactions 25 and 26 repre-sent isomerization reactions. The reactions of ethylene, propyleneand butene on 1,3-butadiene leading to C6-C8 aromatics (reac-tions 34–38) are the equivalent representation of the real reactionsaccounting for the formation of the aromatic products. Reactions 18

5 ering and Processing 49 (2010) 51–58

aolrcht

itrfprCcstab6totftt

magaTrgtsc

t

r

ishna

ctcnd2m3vt

x

1f

Table 3Kinetic parameters for the global decomposition of F-T naphthas.

6 F. Wang et al. / Chemical Engine

nd 31 represent the polymerization reactions for the formationf the liquid products, which account for the decrease of propy-ene and 1-butene during the steam cracking. The formal moleculareaction scheme was proposed as a substitution for the real radi-al addition, decomposition, isomerization, disproportionation andydrogen abstraction reactions in the steam cracking of F–T naph-ha.

The final reaction scheme that contains 38 molecular reactionsncluding the primacy reaction is shown in Table 2. The C4’s frac-ion is an olefin mixture produced from the primary decompositioneaction. In reactions 18 and 31, C5

+ denotes the C6-C8 aromatics-ree product fraction boiling from C5 to the end point of liquidroducts. CnHm denotes the mixture of benzene and toluene. Foreactions 18 and 31, the ratio of the stoichiometric coefficients of5

+ to aromatics was modified to improve the match between thealculated and experimental yields of the C5

+ fraction. The actualtoichiometric coefficients of C5

+ and aromatics in these two reac-ions were determined from the balance of carbon and hydrogennd the experiment results. In the reaction 18, the average car-on number for the CnHm and C5

+ fractions were estimated to be.54 and 6.79, respectively and the average molecular weights wereaken to be 88.28 and 84.99, respectively, on the basis of the balancef carbon and hydrogen and the C5

+ yield in the final products. Inhe reaction 31, the average carbon number for the CnHm and C5

+

ractions were assumed to be 6.54 and 6.79, respectively. Moreover,he average molecular weights for the CnHm and C5

+ fractions wereaken to be 88.28 and 84.99, respectively.

The estimation of the kinetic parameters of the reactions in theolecular schemes is a critical stage in developing a model that

dequately predicts the experimental results. In the reaction 1, thelobal kinetic parameters were determined by a pseudoisothermalpproach through calculating the equivalent reactor volume (VE).he equivalent reactor volume is defined as the one which, at aeference temperature (TR) and a reference pressure (PR), wouldive the same conversion as the actual reactor volume with itsemperature and pressure profiles [24–26]. This concept has beenuccessfully applied to the kinetic analysis and scale-up of steamracking furnaces [11,21,27].

It is well known that the cracking rate of pure hydrocarbons andheir mixtures closely follows a first-order rate law [4,26]:

= A exp(−Ea

RT

)C (1)

The first-order rate constant was also observed in steam crack-ng of naphtha [11,21]. Further, the first-order rate constant washown to be independent of the conversion. The F–T naphtha usederein was a normal paraffinic mixture with few aromatics andaphthenes, so the global reaction was reasonably assumed to befirst-order one.

To determine the kinetic parameters of the reaction 1, a globalonversion was defined. According to the results of analyzinghe feedstock, we defined the global conversion based upon theonsumptions of 18 key components: n-pentane, n-hexane,-heptane, n-octane, n-nonane, n-decane, n-undecane, n-odecane, 2-methylpentane, 2-methylhexane, 2-methylheptane,-methyloctane, 2-methylnonane, 3-methylpentane, 3-ethylhexane, 3-methylheptane, 3-methyloctane and

-methylnonane. It is a weighted mean of the individual con-ersions with respect to the individual mass fractions in the feed,hat is∑

yixi

¯ = ∑yi

(2)

These components, accounting for more than 93.75 wt% in Nap-and 91.79 wt% in Nap-2, are representative of the most abundant

ractions of the naphtha with respect to hydrocarbon structure

Nap-1 Nap-2

Activation energy (kJ/mol) 238.30 222.24Pre-exponential factor (s−1) 1.087 × 1013 2.264 × 1012

and molecular weight. By this method, the dimensionality of mul-ticomponent processes is reduced to one and the chemical andphysical behavior of the naphtha is represented by means of a ficti-tious pseudo-component corresponding to a pseudo-reaction witha first-order rate.

From the mass balance:

F0dx = rdV = rEdVE (3)

the following expression for equivalent reactor volume is obtained[21,26]:

VE =∫ V

0exp(−Ea/RT)dV

exp(−Ea/RTR)(4)

With the expansion and the dilution, the continuity equation forthe F–T naphtha cracking with a first-order rate law may be writtenas

F0dx = k(

1 − x

1 + ı + (ε − 1)x

)(PR

RTR

)dVE (5)

Integration of Eq. (5) leads to

k = F0

VE

(RTR

PR

)∫ x̄

0

1 + ı + (ε − 1)x1 − x

dx (6)

so that the rate constant k may be calculated and the global acti-vation energy (Ea) and the pre-exponential factor (A) are derivedfrom the Arrhenius plots by means of linear regression.

Different experiments at the various cracking conditions wereperformed and reduced to reference temperatures and pressures.Based on Eqs. (4) and (6), a linear regression routine was used tocalculate the values for the global activation energy and the pre-exponential factor. The values obtained for the kinetic parametersof two F–T naphthas are given in Table 3. Since no data are availableon the global activation energies of cracking of such hydrocarbonmixtures in literature, it is not possible to compare these valueswith other ones.

To calculate the kinetic parameters of reactions 2–38 and thestoichiometric coefficients in reaction 1, the objective function wasdefined at outlet of reactor:

˚ =∑

(Y0j − Y)

2

i(7)

where i = 1, . . ., N (N = number of experiments), j = 1, . . ., M(M = number of components), Y0

j= experimental mass fraction of

component j at outlet of reactor, and Yj = predicted mass fraction ofcomponent j at outlet of reactor. The parameters were determinedby substituting the kinetic equations of the model being consid-ered into the appropriate continuity equations for the componentsinvolved in the F–T naphtha pyrolysis. The general form of thecontinuity equations in a tubular reactor with plug-flow can bewritten as

dFj

dl= �d2

in

4Rj = −�d2

in

4

∑i

sijri (8)

The Subplex optimization method [29] was chosen due to its

rapid minimization of the objective function and its satisfactoryperformance in the optimization of kinetic models in this work.The initial kinetic parameters were acquired from the literature[16,21]. All the selected experiments were used for molecular reac-tion system optimization simultaneously, and satisfactory results

ring and Processing 49 (2010) 51–58 57

wef

3

sdahpsetcseaib

stn

caeitraceapNtl

Table 5Comparison between the experimental and simulated yields of products (wt%).

Nap-1 Nap-2

Reference temperature (◦C) 810 810

Severity(C3−/C3H6) 4.6 4.7

Expt.a Simted.b Expt. Simted.

H2 0.74 0.64 0.73 0.62CH4 12.61 12.75 12.97 13.06C2H4 40.78 40.54 39.66 39.37C2H6 5.04 4.93 4.77 4.69C3H6 16.86 17.01 15.9 15.88C4H6 5.20 5.19 5.20 5.621-C4H8 1.81 1.93 1.55 1.70C6H6 1.84 2.16 2.87 2.29∑

C4H8 3.28 3.57 3.03 3.32Light olefins 62.84 62.75 60.76 60.87

F. Wang et al. / Chemical Enginee

ere obtained in the optimization process. The optimized param-ters including the stoichiometric coefficients of primary reactionsor Nap-1 and Nap-2 are shown in Table 2.

.4. Simulation of F–T naphtha steam cracking

A one-dimensional plug-flow model was used to simulate theteam cracking reactor under clean tube condition. Yields of theesired products in olefin production depend on the quality of feeds well as upon the process parameters of cracking furnaces. Sinceeat balance and pressure change continuously during naphthayrolysis, it is necessary to take into account these factors. Sinceteam cracking of F–T naphtha is endothermic, the heat-balancequation is deduced from the fact that the heat transferred throughhe surface of the coil is spent for heating the gas mixture andarrying out the reaction itself. In deriving an equation for the pres-ure variation along the reactor tube, we used the Darcy–Weisbachquation for a straight tube with a round cross-section [18]. Finally,complete mathematical model for the steam cracking process,

ncluding the equations of the mass balance for each species, energyalance and pressure drop, was proposed (Eqs. (8), (9) and (10)):

dT

dl= 1∑N

j=1FjCpj

[U�din(TW − T) + �d2

in

4

∑i

ri(�H)i

](9)

dP

dl= −2�G2E′

�gdin(10)

The integration of this system of differential equations canimultaneously obtain the naphtha conversion, product distribu-ion, and temperature and pressure profiles along the reactor. Theumerical method employed was a Runge-Kutta algorithm.

To evaluate the general accuracy of the model developed, we cal-ulated the average absolute deviations between the experimentalnd simulated values for the main products, including two sets ofxperimental data from Nap-1 and Nap-2. At the present operat-ng conditions, the calculated results were in good agreement withhe experimental data. The average absolute deviations against theepresentative results are tabulated in Table 4. The maximum devi-tion for the individual products is 0.47 wt% for methane in theracking of Nap-1. The average absolute deviations between thexperimental and simulated values indicate that the model has

bility for accurately predicting the ethylene yield for two sam-les. The deviations of propylene are 0.29 wt% in the cracking ofap-1 and 0.21 wt% in the cracking of Nap-2. The results show

hat the light olefin fraction has an average absolute deviation ofess than or equal to 0.54 wt% in all cases investigated. The global

Table 4Average absolute deviations between experimental and simulateddataa.

Component Average absolute deviations (wt.%)

Nap-1 Nap-2

H2 0.08 0.08CH4 0.47 0.16C2H4 0.33 0.26C2H6 0.12 0.11C3H6 0.29 0.21C4H6 0.28 0.251-C4H8 0.19 0.16C6H6 0.27 0.39∑

C4H8 0.34 0.26Light olefins 0.54 0.20BTX 0.55 0.35C5

+ 0.38 0.27

a Simulation conditions: COT, 790–850 ◦C; feedstock flow rate,2.00–2.50 kg/h; and water flow rate, 1.00–1.30 kg/h; and CIP, 1.4 MPa.

BTX 2.68 3.30 3.89 3.40C5

+ 10.44 10.05 11.57 12.10

a Experimental values.b Simulated values.

simulated concentrations of the C6-C8 aromatics and C5+ fractions

agree well with the corresponding experimental values. From theresults shown in Table 4, it can be concluded that the kinetic modeldeveloped can be used to predict the product distribution in steamcracking of naphtha derived from the low-temperature F–T process,as typically presented in Table 5.

4. Conclusions

In this work, steam cracking of F–T naphtha has been carriedout to investigate the characteristics of product distribution andto develop a molecular reaction kinetic model. The experimentalresults indicated that the advantage of the steam cracking of F–Tnaphtha at high severity led to a high ethylene yield and a lowBTX yield. As an excellent feedstock for the steam cracking process,the F–T naphtha presented herein are expected to be a notewor-thy improvement in the production of ethylene. The ethylene yieldfrom the F–T naphtha could exceed 41.83 wt% under an optimumcracking condition. The product yields obtained from the two F–Tnaphtha fractions can be satisfactorily simulated by a molecularreaction model consisting of a first-order primary step togetherwith a set of 37 secondary reactions.

Acknowledgements

The authors gratefully acknowledge the financial support fromNational Natural Science Foundation of China under Grant No.20590361 and the National Outstanding Young Scientists Founda-tion of China under Grant No. 20625620. This work is also supportedby Synfuels CHINA. Co. Ltd.

Appendix A. Nomenclature

List of symbolsF0 molar flow rate of component i (mol/s)Y0

jexperimental mass fraction of component j at outlet ofreactor (wt%)

Yj optimized mass fraction of component j at outlet of reac-tor (wt%)

x fraction conversionyi mass fraction in the feedstockr rate of reaction (mol/(m3 s))rE rate of reaction at reference temperature (mol/(m3 s))V reactor volume (m3)

5 ering a

VkaPPRTTAECFCCdG�RsTU�lE

Gıı�

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

8 F. Wang et al. / Chemical Engine

E equivalent reactor volume (m3)rate coefficient (s−1)stoichiometric coefficientspressure (Pa)

R reference pressure (Pa)gas constant (J/(mol K))temperature (K)

R reference temperature (K)pre-exponential factor (s−1 or m3/(mol s))

a activation energy (kJ/mol)concentration of feedstock (mol/m3)

j molar feed rate of component j (kmol/s)j concentration of component j (kmol/m3)p specific heat of component j (kJ/(kmol K))in internal tube diameter (m)

total mass flow velocity (kg/m2 s)H heat of reaction i (kJ/kmol)

j total consumption rate of component j (kmol/(m3 s))ij stoichiometric coefficient of component j in reaction iW heating temperature at wall of tube (K)

overall heat transfer coefficient (kJ/(m2 K))g density of reacting gas (kg/m3)

axial reactor coordinate (m)′ equivalent coefficient

reek symbolsmolar dilution ratio, mole inert/mole reactantexpansion factor, mole products/mole feedstock crackedfriction coefficient

eferences

[1] A.P. Steynberg, W.U. Nelb, M.A. Desmet, Large scale production of high valuehydrocarbons using Fischer–Tropsch technology, in: 7th Natural Gas Conver-sion Symposium, Dalian, China, 2004, pp. 37–42.

[2] F. Billaud, K. Elyahyaoui, F. Baronnet, Mechanistic modeling of the pyrolysis ofn-hexane, J. Anal. Appl. Pyrol. 19 (1991) 29–40.

[3] S. Nourl, D. Herault, Steam cracking of high-molecular-weight hydrocarbons,Ind. Eng. Chem. Proc. Des. Dev. 20 (1981) 307–313.

[4] R.J. Zou, Q.K. Lou, S.H. Mo, S.B. Feng, Study on a kinetic model of atmosphericgas oil pyrolysis and coke deposition, Ind. Eng. Chem. Res. 32 (1993) 843–

847.

[5] M. Hirato, S. Yoshioka, M. Tanaka, Gas–oil pyrolysis by tubular reactor and itssimulation model of reaction, Hitachi Rev. 20 (1971) 326–334.

[6] L.P. Dancuar, J.F. Maryer, M.J. Tallman, et al., Performance of the Sasol SPDTM

naphtha as steam cracking feedstock, ACS Pet. Chem. Div. Preprints 48 (2003)132–138.

[

[

[

nd Processing 49 (2010) 51–58

[7] M.E. Masoumia, S.M. Sadramelia, J. Towfighia, A. Niaei, Simulation, optimizationand control of a thermal cracking furnace, Energy 31 (2006) 516–527.

[8] E.H. Edwin, J.G. Balchen, Dynamic optimization and production planning ofthermal cracking operation, Chem. Eng. Sci. 56 (2001) 989–997.

[9] W.R. Shu, L.L. Ross, Cracking severity index in pyrolysis of petroleum fractions,Ind. Eng. Chem. Proc. Des. Dev. 21 (1982) 371–377.

10] M. Dente, E. Ranzi, A.G. Goossens, Detail prediction of olefin yields from hydro-carbon pyrolysis through a fundamental simulation model (SPYRO), Comput.Chem. Eng. 3 (1979) 61–75.

11] K.M. Sundaram, G.F. Froment, Modeling of thermal cracking kinetics. 3. Radicalmechanisms for the pyrolysis of simple paraffins, olefins, and their mixtures,Ind. Eng. Chem. Fundam. 17 (1978) 174–182.

12] D.S. Aribike, A.A. Susu, Mechanistic modeling of the pyrolysis of n-heptane,Thermochim. Acta 127 (1988) 259–273.

13] E. Joo, S. Park, M. Lee, Pyrolysis reaction mechanism for industrial naphthacracking furnaces, Ind. Eng. Chem. Res. 40 (2001) 2409–2415.

14] D. Depeyre, C. Flicoteaux, F. Arbabzadeh, A. Zabaniotouf, Modeling of ther-mal steam cracking of an atmospheric gas oil, Ind. Eng. Chem. Res. 28 (1989)967–976.

15] K.M. Sundaram, G.F. Froment, Modeling of thermal cracking kinetics. I. Ther-mal cracking of ethane, propane and their mixtures, Chem. Eng. Sci. 36 (1977)601–608.

16] K.M. Sundaram, G.F. Froment, Modeling of thermal cracking kinetics. II. Crack-ing of iso-butane, of n-butane and of mixtures ethane-propane-n-butane,Chem. Eng. Sci. 30 (1977) 609–617.

17] M. Murata, S. Saito, A simulation model for high-conversion pyrolysis of normalparaffinic hydrocarbons, J. Chem. Eng. Jpn. 8 (1975) 39–45.

18] A.M. Aliev, A.Z. Tairov, A.M. Guseinova, Y.M. Kalaushina, T.N. Shakhtakhtinskii,Optimal design of alkane pyrolysis processes: propane pyrolysis, Theor. Found.Chem. Eng. 38 (2004) 654–659.

19] D.S. Aribike, A.A. Susu, A.F. Ogunye, Mechanistic and mathematical modelingof the thermal decomposition of cyclohexane, Thermochim. Acta 51 (1981)113–127.

20] P.S. Van Damme, G.F. Froment, Scaling up of naphtha cracking coils, Ind. Eng.Chem. Proc. Des. Dev. 20 (1981) 366–376.

21] P. Kumar, D. Kunzru, Modeling of naphtha pyrolysis, Ind. Eng. Chem. Proc. Des.Dev. 24 (1985) 774–782.

22] X.W. Zhao, Y.L. He, G.Z. Wang, J.Y. Bao, Pyrolysis laboratory unit and its appli-cation, Petrochem. Technol. 32 (2003) 70–72.

23] E. Van Camp, P.S. Van Damme, P.A. Willems, P.J. Clymans, G.F. Froment, Severityin the pyrolysis of petroleum fractions. Fundamentals and industrial applica-tion, Ind. Eng. Chem. Proc. Des. Dev. 24 (1985) 561–570.

24] G.F. Froment, O. Boudewijn, V.D. Steene, J. Philip, V. Berghe, A.G. Goossens,Thermal cracking of light hydrocarbons and their mixtures, AIChE J. 23 (1977)93–106.

25] G.F. Froment, O. Boudewijn, V.D. Steene, P.S. Van Damme, S. Narayanan, A.G.Goossens, Thermal cracking of ethane and ethane–propane mixtures, Ind. Eng.Chem. Proc. Des. Dev. 15 (1976) 495–504.

26] P.S. Van Damme, S. Narayanan, G.F. Froment, Thermal cracking of propane andpropane–propylene mixtures: pilot plant versus industrial data, AIChE J. 21(1975) 1065–1073.

27] C.E. Van Camp, P.S. Van Damme, G.F. Froment, Thermal cracking of kerosene,Ind. Eng. Chem. Proc. Des. Dev. 23 (1984) 155–162.

28] K.K. Pant, D. Kunzru, Pyrolysis of n-heptane: kinetics and modeling, J. Anal.Appl. Pyrol. 36 (1996) 103–120.

29] T. Rowan, Functional stability analysis of numerical algorithms, Ph.D. thesis,University of Texas at Austin, 1990.